Recursive Variational Mode Decomposition Algorithm for Real ...www.researchgate.net › publication › fulltext › Recursive-

ScienceDirect Procedia Technology 21 (2015) 540 – 546

SMART GRID Technologies, August 6-8, 2015

Recursive Variational Mode Decomposition Algorithm for Real Time Power Signal Decomposition Soman K Pa, Prabaharan Poornachandranb, Athira Sa*, Harikumar Ka a

Center for excellence in Computational Engineering and Networking, Amrita Vishwa Vidyapeetham b Amrita Centre for Cyber Security systems and Networks, Amrita Vishwa Vidyapeetham

Abstract Conventional methods of signal decomposition are observed to fail in power system applications and computationally intensive algorithms like EMD, VMD, EWT are found to give better performance. The heavy computations associated with them restricts their use in real time applications and stream processing. This paper presents a recursive block processing technique for real time signal decomposition. The use of recursive FFT and the clever initializations of the center frequencies in the existing VMD algorithm helps in reducing the computational complexity and hence speeds up the process. This low complexity algorithm was tested on synthetically generated power signals and the results were observed to be consistent with the existing VMD algorithm. © 2015 2015Published The Authors. Published by Elsevier Ltd. by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of Amrita School of Engineering, Amrita Vishwa Vidyapeetham University. Peer-review under responsibility of Amrita School of Engineering, Amrita Vishwa Vidyapeetham University Keywords:Recursive FFT, VMD, computational complexity, power signal, decomposition.

1. Introduction The advent of smart grid technology and the modern day electronics has introduced many sensitive devices to our electrical distribution networks and this makes the maintenance of power quality a critical issue. Deviations from the standard power signal parameters results in poor device performance and may even lead to permanent damages in certain situations. Hence, there is the need for efficient real time power quality monitoring systems which can identify even slight variations in the signal parameters and thus take preventive action to avoid damages due to abnormal system behaviour. Power quality is typically a function of power signal parameters and hence a signal processing technique to identify signal variations can effectively identify distortions in the power signal. Signal

* Corresponding author. Tel.09790446239 E-mail address:[email protected]

2212-0173 © 2015 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of Amrita School of Engineering, Amrita Vishwa Vidyapeetham University doi:10.1016/j.protcy.2015.10.048

K.P. Soman et al. / Procedia Technology 21 (2015) 540 – 546

analysis and inference is a widely explored field of signal processing. One of the oldest and simplest methods for the frequency analysis of a stationary signal is the analysis of the signals Fourier spectra, however this method essentially fails in the case of power signals with distortions which are non-stationary in nature. Methods like the short time fourier transform (STFT) analyses the signal over fixed sized windows over which the signal characteristics are assumed to be stationary and hence offers good time resolution but poor frequency resolution (uncertainty principle). The Fast Fourier Transform (FFT) introduced formerly by Gauss came to be used widely after the paper by Cooley and Tukey [9] in 1965 accelerated the heavy computations behind Fourier transform and made the real time implementation of these algorithms imaginable. Then came up the wavelet transform which uses size adjustable windows for analysis and thus offers better resolution as compared to methods like STFT and also offers better denoising methodologies. However the performance heavily relies on the choice of wavelet and the number of decomposition levels chosen for analysis. [2], [3] discusses methods for implementation of algorithms using wavelet transforms in real time. All these methods decomposes the signal into a set of fixed basis and hence proves to be the wrong choice of decomposition technique in many instances. Further, they may lead to misleading inferences in many cases due to their poor adaptability to the signal nature. Basis functions which could adapt to signal nature would thus be a better choice for representation of all types of signals. It was with this idea that the Empirical Mode Decomposition was proposed by Huang et.al [11], and gave satisfactory results in certain cases where the conventional methods failed; but there was no strong mathematical reasoning behind the framework. [4], [5] and many more works tried to develop a mathematical reasoning for the same and arrived at similar techniques with slight variations and different interpretations. One such work by Daubechies [10] resulted in the synchrosqueezed wavelet transform which proved t